AI Companies Are Hiring Improv Actors to Train Chatbots to Understand Human Emotion

Every year, AI systems get better and better, but they still have trouble understanding how people feel and how to have a natural conversation. Modern AI models can write essays, answer questions, and make pictures, but they often miss the small emotional cues that make real-life interactions happen.

Some AI companies are now turning to an unexpected group of experts to help solve this problem: improv actors and performers. These people are being hired to help train AI systems by having real, unplanned conversations that models can learn from.

At first, the idea might sound strange, but more and more AI developers think that machines need more realistic data about how people interact with each other in order to learn how to communicate better.

Why AI Still Has Trouble with Emotion

Over the past few years, large language models have gotten a lot better. They can write text that sounds natural and convincing, but they still don't understand how emotions work.

It's not just words that make up human communication. It has:

• How you say it

• Pauses and timing

• Facial expressions

• Context of emotion

• Language of the body

AI systems have a hard time understanding these things because they need to understand social cues in a more complex way than just recognizing patterns.

Researchers often say that AI systems are "jagged" because they can do very well at difficult tasks but not at simple social interactions.

For instance, a chatbot might give technically correct answers but not get sarcasm, humor, or emotional tone. This gap is even more obvious as businesses make voice-based AI assistants that are supposed to talk to people in a natural way.

What Improv Actors Do to Help Train AI

Companies are trying out a new way of training AI models to help them better understand and show emotions.

Handshake AI, a company that provides training data to big AI labs, has posted job openings for actors, comedians, and improvisers to take part in group training sessions.

Instead of acting out scripted scenes, participants are asked to do improv exercises over video calls.

In these sessions:

• People who perform are given prompts or situations

• They make up conversations and interactions

• They change their moods or personalities

• In real time, they respond to each other in a natural way
  

Then, these interactions are recorded and studied to help teach AI models how people show their feelings when they talk to each other.

The job description says that the goal is to make interactions that seem “real and human.”

Improv actors are especially good at this job because they are trained to react on the spot and look into emotional dynamics as they happen.

The Increasing Need for Human Training Data

Using actors to teach AI is a sign of a bigger trend in the tech world.

Companies are racing to make AI systems that are more powerful, and they are finding that high-quality human data is one of the best ways to make models better.

To train modern AI systems, you need a lot of data. But raw internet data doesn't always have the structured examples that machines need to learn how to do things like reason, be aware of their feelings, or be an expert in their field.

This has led to the growth of a new business that focuses on training data made by people.

Companies now hire people from a wide range of fields to help train AI systems, such as:

• Medical researchers and doctors

• Lawyers and people who study the law

• Engineers of software

• Writers and reporters

• Actors and other performers
  

These experts give examples, look at AI answers, and help make realistic situations that models can learn from.

In a lot of cases, these workers are actually teaching AI systems how to do their own jobs.

How Much Do These Jobs Pay?

Handshake AI's improv training jobs pay an average of $74 per hour, which makes them appealing to many actors who want to work on their own schedule.

The job postings say that the work is part-time and can be done from home, so actors can do it while they are auditioning, rehearsing, or working on other creative projects.

Some workers have said, though, that the initial pay rates can go down after onboarding and that tasks aren't always available.

Workers may have to compete for a small number of jobs because these projects usually take place online.

This is part of a larger trend in the new AI gig economy, where a lot of contractors do short-term jobs that help train and improve machine-learning systems.

Worries from the Creative Community

Some people don't like the idea of actors helping to train AI.

Some performers are worried that by sharing their talents, they might be helping to make technology that could eventually take their jobs.

If AI systems can write realistic dialogue, act out emotions, or even make fully animated characters, they could make it less necessary for people to act in some fields.

These worries have sparked discussion in online groups. Some improv performers have said that the trend is disturbing or even dystopian in discussions.

Some people have joked about how ironic it is that AI is going after improv comedy, which is already known for being financially unstable.

Some performers, on the other hand, see the chance as a new way to make money while helping new technologies grow.

The Growth of Multimodal AI

The push to make AI better at understanding emotions is closely related to the growth of multimodal AI systems.

Multimodal models are made to talk to users in more than one way, such as:

• Text

• Voice

• Pictures

• Video
  

A lot of AI companies are now making voice assistants that can talk to you in a way that sounds real and has real emotion.

These systems need training data that shows how people naturally talk to each other in order to work well.

That's one reason why businesses are hiring actors and other performers. Their ability to improvise gives them the kind of expressive and spontaneous interaction that machines have a hard time copying.

The Future of Working Together with AI

Using improv actors to teach AI shows that AI still depends a lot on human knowledge, even though technology is moving quickly.

People give machines examples, feedback, and emotional context that help them get better.

This kind of work together could happen more often in the future. People from many fields, including artists, teachers, scientists, and professionals, could help shape how AI systems learn.

But this also brings up tough questions about who owns what, workers' rights, and how AI will affect creative jobs in the long run.

If intelligent machines learn from human creativity, the way people and technology work together could change a lot.

Final Thoughts

AI companies hiring improv actors shows how hard it is to teach machines about the many different ways people feel.

AI models are very good at writing text and fixing technical problems, but they still have trouble with the small things that make up human interaction.

Actors can help close that gap and make AI systems better at talking to people by recording real conversations and emotional exchanges.

The trend also shows that there is more tension in the AI economy. The people who teach these systems might also be working on technologies that could change or even destroy their own jobs.

The partnership between human creativity and machine intelligence is likely to become one of the most important stories in the tech industry as AI keeps getting better.